Anti-online harassment session at CES 2016

Above: Anti-online harassment session at CES 2016

Image Credit: Dean Takahashi

Question: What do you think of Intel’s push into diversity? They made a bigger deal about that at the last two keynotes.

Su: It’s an admirable effort. They’ve done a good job with it, putting the focus on it. Everyone in the semiconductor industry, everyone in the technology industry, would benefit from more diversity in the business.

Question: When you go into OEMs or other potential customers and talk to them, what are they looking for? What do they ask for?

Su: We have jointly picked the design point on Zen with our top customers being very involved. From a design standpoint, I think they’re very pleased with it. Our customers want Zen to be in the market. The focus is very much on co-development. We’re doing the chip development as they do the platform development so we can enhance time to market. There’s a huge desire to see Zen in the marketplace. That’s true particularly on servers, but also on the PC side as far as high-performance desktops. We’re very committed to doing that. We’re enabling them with strong samples, platforms, and software availability.

The channel ecosystem, as you guys can imagine, is very fired up about the platform. I get a lot of notes from enthusiasts saying that they’d like to see Zen sooner. It’s cute.

Question: The server market has changed in the last few years. It’s fragmented into very different kinds of servers. Are there particular segments where you think Zen is going to be targeted within that broader market?

Su: We’ll do very well in multithreaded workloads and workloads that benefit from a lot of parallelism. When you think about some of the cloud workloads that do that, we’ll target those segments. We also believe there are opportunities to grow the one-processor server base as well. We do have a good sub-segmentation of where things go. Our overall focus is trying to get those highly multithreaded workloads an advantage.

Question: Given the strength of your partnerships — you look across the chip industry at the likes of ARM, who don’t make a big push at putting their brand on everything. Are we seeing a decline of the need to be seen as being in something? Or does that still have value? Do people want to know that there’s an AMD chip in something? Or is that less important now?

Su: It’s important to be there, for sure. It’s important to be viewed as a technology leader. That doesn’t necessarily mean AMD branding on the box. But that means, if we’re in five of the top six applications, that’s a pretty good view. That’s really our focus. How do we get into the top brands? For instance, being Apple’s discrete graphics partner across the Mac line, that’s something we’re very proud of. We’re never going to be on the front of the box, but that’s a marquee application. That’s the way we think about it.

Question: Why do you think we saw so many big deals in chips this past year? Everything was in a steady state for the longest time, and now we’re seeing $37 billion takeovers.

Su: I do think that says a bit about the maturity of the semiconductor industry. When you look at the industry and how you can extract profitability growth, you either do it through innovation or you do it through efficiency. The M&As help on that efficiency aspect. There’s a lot to do on the innovation aspect as well.

Microsoft Surface pro 4 2

Above: Microsoft Surface Pro 4.

Image Credit: Microsoft

Question: If you look back over the last few years, clearly the pace of change in terms of semiconductor process technology has slowed down. As a customer of all these guys, does the increasing time between nodes, the increasing cost of each of those nodes, does this bother you? How do you think of this as a customer of the big chipmakers?

Su: In some sense it’s an equalizing force. If you look at it, back five or seven or eight years ago, you’d see that process technology, are you first into a process technology? That ended up being one of the bigger differentiators of companies. As we go forward, 28nm was a really long node. That made architectural improvements and design improvements much more important. The FinFET 16nm or 14nm will be a very long node with a lot of products. That affords much more architectural optimization.

I also look at the gap between us and our competitors. The idea of the process technology lead being the sole and determining factor of who wins really does change the game. It will be much more about architecture and design. It’s a great equalizer from that standpoint going forward.

Question: The only counterexample off the bat is just the power aspect of at all. Architecture and design is one thing. That’s something AMD has had a leadership role in dating back to the K6. But over time, because of power becoming more of an issue, process technology is the easy way to go. If you have a more advanced process generation than your competitor, you’re going to be lower power as well.

Su: There’s no question that process technology is a big knob to turn in the performance and power spectrum. That’s why we’re going FinFET relatively early. We saw the power benefit. My comment is relative to where we see it going long term. We’ll always be in advanced technology. My view is that 16nm and 14nm are very important. We’re transitioning a number of products there in 2016. But it’ll be a long node. It’ll last for the next three, four, five years.

Within that node we can do a lot of optimization. Even on power there’s a lot we can do in the architecture as far as power management, power tuning, the power optimization between hardware and software. Once you’re in the node it becomes all about architecture. Every three, four, five years you make one big jump with process technology and then the rest is on architecture.

Question: I have a Microsoft Surface Book with an Intel chip inside here. As we go to more of a two-in-one environment, how does AMD play into that? Can you keep up with Intel in that space?

Su: Absolutely. Particularly with this node, as we go into FinFET. The power efficiency of the node is very significant. We’ll see a significant improvement. We’ve made improvements in our APU power efficiency already through design and some of the things we’ll announce later this year. As we go to the seventh-generation APU you’ll see big improvements. But certainly the process technology will help in that range.

Question: In the end, in the market, out of 5,000 notebooks in the European market, there are only 25 with Carrizo, and only three of them have an SSD.

Su: The Carrizo APU, from a technology standpoint, is outstanding. If you look at the performance and performance per watt improvements, outstanding. When you look at the OEM systems in the marketplace, we’ve made good progress. You look at HP’s EliteBooks, those are nice machines.

I think people need to see existence proof. The idea of AMD being the cheap solution has to be replaced by AMD as a very competitive solution. The HP notebooks out there are one example. The Lenovo gaming system is another very nice set of machines. Dell has a number of machines out that are strong. You’ll continue to see that portfolio increase. Each generation will get better.

If you looked at a year ago, you probably saw nothing at the $1,000 price point. Now you see three. The next generation, hopefully it’ll go up a little faster.